方堃 Kun FANG

方堃 Kun FANG

Postdoctoral Fellow

About

I am a Postdoctoral Fellow at ASTAPLE, where I am honored to work with Prof. Haibo HU, in Department of Electrical and Electronic Engineering of The Hong Kong Polytechnic University.

My research aims at trustworthy deep learning for models deployed in open-world environments, where inputs may fall outside the training distribution and training data may need to be removed on request. I pursue this through two concrete tasks that address the robustness and privacy of these models: out-of-distribution detection and machine unlearning.

Trustworthy Deep Learning Open-World Learning Out-of-Distribution Detection Machine Unlearning

Research Interests

Out-of-Distribution Detection

Can the pretrained model recognize unseen and distribution-shift inputs?

#robustness · #generalization

Machine Unlearning

Can the pretrained model forget specific knowledge of seen training samples?

#privacy
A common thread: low-dimensional structure. Both problems can be approached through a low-dimensional view of data representations and model parameters — a principled lens for analyzing, detecting, and manipulating what makes a model trustworthy.

Earlier work also touched on kernel methods and adversarial robustness.

Education & Experience

  • 2025.04 – now Postdoc The Hong Kong Polytechnic University ASTAPLE · Advisor: Prof. Haibo HU
  • 2021.04 – 2025.03 Ph.D. Shanghai Jiao Tong University Advisor: Prof. Jie YANG
  • 2018.09 – 2021.03 Master Shanghai Jiao Tong University
  • 2014.09 – 2018.06 Bachelor Tongji University

Honors & Awards

  • Top Reviewer — NeurIPS 2026
  • Gold Reviewer — ICML 2026
  • Outstanding Reviewer — CVPR 2026 & ECCV 2024

Last updated on 2026.10.30 by DeepSeek